This action might not be possible to undo. Are you sure you want to continue?

DMTCS vol. 12:3, 2010, 63–74

**´ Covering codes in Sierpinski graphs
**

Laurent Beaudou1

1 2

Sylvain Gravier2

Sandi Klavˇ zar3,4,5

Matjaˇ z Kovˇ se2,5

Michel Mollard2

LaBRI - Universit´ e Bordeaux 1 - CNRS, 351 cours de la Lib´ eration, 33405 Talence cedex, France. Institut Fourier, UJF - CNRS, 100, rue des Maths, BP 74, 38402 St Martin d’H` eres Cedex, France. 3 Faculty of Mathematics and Physics, University of Ljubljana, Jadranska 19, 1000 Ljubljana, Slovenia. 4 Faculty of Natural Sciences and Mathematics, University of Maribor, Koroˇ ska 160, 2000 Maribor, Slovenia. 5 Institute of Mathematics, Physics and Mechanics, Jadranska 19, Ljubljana, Slovenia.

**received 4th February 2009, accepted 14th May 2010.
**

For a graph G and integers a and b, an (a, b)-code of G is a set C of vertices such that any vertex from C has exactly a neighbors in C and any vertex not in C has exactly b neighbors in C . In this paper we classify integers a and b for which there exist (a, b)-codes in Sierpi´ nski graphs. Keywords: codes in graphs, perfect codes, Sierpi´ nski graphs

1

Introduction

In coding theory, a binary code is deﬁned as a subset of {0, 1}n . Since a code should at least correct one error, the Hamming distance between any pair of code vertices must be at least 3. In terms of graph theory, we seek for a vertex subset C of the n-cube Qn , such that d(u, v ) ≥ 3 for any u, v ∈ V (Qn ), u = v . (As usual, d(u, v ), is the shortest-path distance between u and v .) The above concepts can be extended from hypercubes to arbitrary graphs. Let G be an arbitrary (connected) graph. Then a subset C of vertices of G is called a 1-code (or simply a code) if d(u, v ) ≥ 3 holds for any u, v ∈ C , u = v . Moreover, C is a perfect code provided that the closed neighborhoods of elements of C form a partition of V . It was Biggs [2] who initiated the study of perfect codes in distance regular graphs. Kratochv´ ıl with his co-workers follows with the study of codes in general graphs, see the monograph [16], references therein, and [9] for result on the related complexity issues. Tower of Hanoi graphs model the classical Tower of Hanoi puzzle with 3 pegs and n discs. Their nice fractal structure enables to observe many nice properties. In particular, Cull and Nelson [5] proved that they contain (essentially) unique perfect codes, see also [17]. The Tower of Hanoi graphs extend naturally to graphs S (n, k ), n, k ≥ 1, where S (n, 3) are isomorphic to the graphs of the puzzle with 3 pegs and n discs. The theorem of Cull and Nelson was extended to S (n, k ) in [14], see also [8] where in particular shorter arguments are provided.

c 2010 Discrete Mathematics and Theoretical Computer Science (DMTCS), Nancy, France 1365–8050

k ) (n. . j2 . therefore completing the proof of the Theorem 1. Cohen. Honkala. an (a. b)-codes in graphs S (n. (a. k ). b)-code. 2. (a. . b = a + 1. in ) and v = (j1 . (v) a ≥ 1. k ) contains an (a. The motivation for their introduction were topological studies from [19. Let G be a graph and a. k ). b)-code of G if any vertex from C has exactly a neighbors in C and any vertex from G \ C has exactly b neighbors in C . Matjaˇ z Kovˇ se. Finally. n ≥ 1. form a two-parametric family of graphs of fractal type and have been wellstudied by now. j )-coverings from [1] are exactly (i − 1. n} such that . i. Theorem 1. 2. We will simply speak about (a. The graph S (n. Sandi Klavˇ zar. are of no special interest. Telle deﬁnes [i.1 Let n. 2 Graphs S (n. we have already mentioned perfect codes. Then a set C of vertices of G is an (a. . Independently. k ≥ 2. . k odd. j )-coverings. Axenovich [1] studied some special cases of perfect weighted codes calling them (t. . k ) admits an (a. (iii) a = 0. Since S (n. k ) see the recent Lipscomb’s book [18]. k ) were introduced in [13] and later named after Sierpi´ nski in [14]. For this aspect of the graphs S (n. Sylvain Gravier.) In this paper we give a characterization of the parameters n and k for which S (n. We proceed as follow. a). let us assume in the rest that n ≥ 2 and k ≥ 2. k = 2a + 1. Lytsin and Mattson introduced a generalization of covering codes using weights and named them weighted codes. 1)-labelings [8]. k ≥ 1. b = 1. k ). n odd. b)-codes when referring to such sets. and different colorings [12. b )-covering as deﬁned in [4]. j )-codes. in Section 3. . Moreover. (iv) a ≥ 1. For a study of small radius weighted coverings see [3]. k ≥ 1) is deﬁned on the vertex set {0. Indeed. . 20]. We conclude with some ideas for further research. j ]-dominating sets in [21] which are exactly (i. . Michel Mollard In [4]. b)-code is exactly a perfect a 1 ( b− b . (See the next section for their deﬁnition and basic properties. two different vertices u = (i1 . b = a. b = a. . (1. and S (1. jn ) being adjacent if and only if there exists an index h in {1. if two vertices from a (0. Note that (0. Graphs S (n. Then. crossing numbers [15]. . j )-codes. a + 2). i2 . Other aspects include L(2. and (a.64 Laurent Beaudou. (ii) a ≥ 2 even. k ) were studied from many different points of view. b)-code if and only if a < k and one of the following cases holds: (i) a ≥ 1. 1)-code would be at distance 2. Deﬁned in this way. b = a + 2. 1). i. Hence Theorem 1. then their common neighbor would have two neighbors in the code. . see also [7]. k even. In the next section we introduce and describe graphs S (n. a + 1)-codes. .1 (iii) covers the before mentioned result on perfect codes in graphs S (n. . b nonnegative integers. 11]. Then S (n. 1. Then our main result is the following. In the subsequent sections we construct the claimed (a. . . k − 1}n . k ) Graphs S (n.1. k ). we give necessary conditions on a and b for the existence of (a. . where these graphs are addressed as Klavˇ zar-Milutinovi´ c graphs. n = 2. b)-codes were introduced in the following way [6]. k ). 1)-codes coincide with perfect codes. k ≥ 1. n. In order to have a more convenient deﬁnition of perfect weighted coverings of radius one. Graphs S (n.

. for t = 1. 3) Note that the graph S (n. Note that the extreme vertices of S (n. S (n. 1) = K1 . . if n = 1. . . b) codes for them are (0. 1 ≤ j ≤ k . 1)-code for any n ≥ 1. On the other hand. . 3). . The set of extreme vertices of S (n. S (2. and S (3. 2)-code if and only if n is odd. P2n has a (0. then S (1. . In the latter case. has only one vertex. A vertex of the form ii . S (3. k ) = Kk . 3). . k ))| = k and that |S (n. k ) are of degree k − 1 and the degree of the inner vertices is k . 4). k ) will be denoted X (S (n. i2 . i of S (n. (1. . 1. . k )| = k n . 1)-code and a (1. h − 1. More precisely. k )). . Note that S (n. in as short for (i1 . Finally. (ii) ih = jh . . . 4) are shown in Fig. Note also that |X (S (n. while it contains a (1. n. . 2) is isomorphic to the path on 2n vertices. 3 Necessary conditions In this section we restrict the possible couples (a. k ) are possible. i n− 1 j . for a vertex u = i1 i2 . . 3). 1) and (1. Fig. . 65 In the rest of the paper we will write i1 i2 . . 1: S (2. k ) is called extreme. . where . . The graphs S (2. n ≥ 1. and (iii) it = jh and jt = ih for t = h + 1. other vertices are inner. b)-codes in graphs S (n. in of S (n.Covering codes in Sierpi´ nski graphs (i) it = jt . b) for which (a. 2). 1). in ). The only possible (a. k ) let K (u) denote the k -clique induced by vertices i 1 i 2 . and S (2. 1). . such a code is 1 101 101 · · · 101 1. that is.

Suppose ﬁrst that there exists u ∈ X (S (n. Let v be a vertex in X (S (n. 2)-codes for odd n. y has degree k . or Kk ⊂ C and |C ∩ Kk | = k > b − 1. as shown at the beginning of the proof.1 Let C = ∅ be an (a. k ) is not regular and thus. Other vertices of V1 have degree k − a. which yields a contradiction. We now assume a > 0. We will therefore assume in the rest that k ≥ 3.3 Let C be an (a. k ) with bipartition V1 = C and V2 = V (S (n. keeping only the edges between V1 and V2 . Since n ≥ 2. Then |C | · (k − a + b) = bk n + d . Since n ≥ 2.5 Let C be an (a. Proof: If there is a clique Kk with |C ∩ Kk | > a +1. b)-code in S (n. Then it follows that C = V (S (n. 0)-code nor a (k − 1. which is not possible. Consider the clique K (u). Then u is adjacent to b vertices in K (u). one more in C .4 Let C be an (a. y cannot have a neighbors in the code and y is not in the code. Matjaˇ z Kovˇ se. b)-code of S (n. Consider a clique Kk . Sandi Klavˇ zar. b)-code of S (n. Then u has at most one neighbor not in K (u) and |C ∩ K (u)| ≥ b − 1 so that b ≤ k . in particular. Then counting the number of edges in B in two ways we have |C | · ( k − a) − d = (k n − | C | ) · b . S (n. k ) without extreme vertices. Then a < k and b > 0. Consider y a neighbor of x. Proof: Consider the bipartite subgraph B of S (n.66 Laurent Beaudou. k ) and Kk any of its k -cliques. An arbitrary code vertex from K (u) has b − 1 neighbors in K (u) ∩ C and. which is neither a (k. ✷ Lemma 3. b)-code in S (n.2 Let C be an (a. k )). Since x has degree k − 1. Then b − 1 ≤ | C ∩ Kk | ≤ a + 1 . Therefore b > 0. the statement holds trivially. Either there is a vertex v in Kk \ C and the preceding study yields |C ∩ Kk | ≥ b − 1. Then its degree in B is k − a − 1. Every vertex in V2 has degree b in B . k ) with d extreme vertices in C . k )) \ C . Suppose a = k and let x ∈ X (S (n. Lemma 3. Lemma 3. and at least one of them is not in the code (namely x). b)-code of S (n. There exists a vertex u which is not in C . Sylvain Gravier. ✷ . Therefore. k ). Michel Mollard 1 denotes code vertices. Then (k − a + b)|bk n . maybe. ✷ Lemma 3. k )) ∩ V1 . from which the lemma follows. item (iv) is covered with the (1. Proof: Suppose b = 0. k )) \ C . none of the neighbors of x is in the code and b = 0. Thus. then any vertex v ∈ C ∩ Kk satisﬁes |N (v ) ∩ C | > a. Clearly a ≤ k . the code cannot be the whole graph. Proof: If a = 0. Hence Theorem 1. k ). 0)-code. Thus a ≤ b. it is not in the code. Corollary 3. k )). Then a ≤ b.1 holds for k = 2.

Let v be another neighbor of u (we recall that k ≥ 3). then n = 2 and k = 2a + 1. Hence x is adjacent precisely to b − 1 code vertices in K (x). a + 1). Since y and x have already one neighbor in the code. k ) with a odd. Since a − 1 > 0 we can consider such a vertex.7 If an (a.2. a)-code in S (n. a + 2). Then z ∈ / C . does not belong to C . Else. it means that K (u) ⊂ C . Either it is in C and a > 1 or it is not in C and has at least two neighbors in the code so that a ≥ 2. we get |K (u) ∩ C | = a − 1. Let z be the neighbor of y that is not in K (x). a)-code for any a ≥ 1. it means that |C | must be even. Now. If b > 1. such a vertex exists since a > 0. |K (u) ∩ C | = a. Suppose ﬁrst that there is some x ∈ X (S (n. b)-codes are (a. and k odd. It follows that there is no extreme vertex in C . (a. By way of contradiction.2. Thus the above implies that |C | = a · k n−1 . we may state that K (x) and K (y ) contain no vertex of the code. k )) ∩ C . Since K (u) contains u and w. ✷ Note that it follows from the above proof that no extreme vertex can belong to an (a. k ). by Lemma 3. Since we have assumed that k ≥ 3. a). From Lemmas 3. Next we will obtain some additional necessary conditions for the existence of such codes. for otherwise y would be adjacent to more code vertices than x. b = 1. a contradiction. a ≥ 1. |C ∩ Kk | = a + 1 for any k -clique Kk . Since k ≥ 3. suppose that there is a vertex u such that |K (u) ∩ C | < a. We have thus shown that for every vertex u. it is in the code and its neighbor y not in K (u) is not for the same reason as x. Thus |C | = k n−1 · (a + 1) . z has a − 1 neighbors from C . Then x ∈ K (u) ∩ C is adjacent to at most a − 1 code vertices. Lemma 3. Then. a)-code in S (n. for otherwise w would be adjacent to more code vertices than u.5 we deduce that the only possible (a. . Then the neighbor x of w. k )) ⊂ C . the number of k -cliques in S (n. Lemma 3. However. Consider vertices x and y linking these two cliques.2 and 3. Therefore any clique Kk contains either 0.5 we need k > 2 when we study the case b = 1. each code vertex of K (x) has at most one other neighbor in C \ K (x). On the other hand. in K (z ). there exists another vertex v in K (x). y = x. a + 2)-code exists in S (n. Thus a ≤ (b − 2) + 1 = b − 1. we have assumed in the beginning of the section that this is indeed the case. Lemma 3. Thus we may conclude that b > 1. ✷ Note that in the proof of Lemma 3.6 If a and k are odd then there is no (a. Let w = u be a vertex in C ∩ K (u). Since a is odd. Let u be an extreme vertex of S (n. Proof: By Lemma 3. k ). k ) is k n−1 . k ). |K (u) ∩ C | = a. and (a. 1 or k vertices of the code. Because the subgraph induced by C is a-regular with a odd. It can have at most a − 1 adjacent code vertices. suppose X (S (n.3 implies that |C | · (k − a + (a + 2)) = k n−1 · (a + 1) · (k + 2) = k n · (a + 2) + d . We next claim that for any vertex u. Proof: Let C be an (a. x ∈ / K (u).Covering codes in Sierpi´ nski graphs 67 Now. there exists a vertex y ∈ K (x) ∩ C . But a · k n−1 is odd. They have no neighbor in the code which is impossible.

Moreover ij is also adjacent to ji ∈ C . . . Second. it follows that p < k and vertex ij has precisely 2p − 1 neighbors in K (ij ) ∩ C . a + 2)-codes. Matjaˇ z Kovˇ se. k ) contains an (a. a)-code in S (2. Let n ≥ 3 and suppose that S (n − 1. Therefore ij has exactly 2p = a neighbors in C . in ∈ C. k ) exists if and only if (i) a is even or (ii) a is odd and k is even. It is now straightforward to see that C is an (a. Hence by Lemma 4. k k n− 2 By the right-hand side of the last equality we infer that this number is between 0 and 1 (recall that n ≥ 2).1 Suppose there exists an (a. ✷ Lemma 4. b)-code C that does not include any of its extreme vertices. 0 ≤ i ≤ k − 1. a) and (a. Sandi Klavˇ zar. in ∈ C if and only if i2 . Proof: Suppose S (2.6. a)-code exists. if i = j it follows that (j − i) (mod k ) > p. Case 1. Sylvain Gravier. since ij ∈ / C . a). Dividing both sides of the equality by k n−1 we arrive at 2a + 2 − k = d 1 · .2 Let n ≥ 2 and a < k . . b)-code in S (2.1 it sufﬁces to prove the existence of codes for the graphs S (2. a and k cannot both be odd if an (a. Let ij ∈ C . Michel Mollard where 0 ≤ d ≤ k . thus it can only be 1 or 0. ✷ 4 Existence results for (a. we have observed that no extreme vertex belongs to such a code. Then an (a. Lemma 4. 1 ≤ ε ≤ p} . where ε = ε. k ) contains an (a. b)-codes in S (n. a = 2p. 1 ≤ ε ≤ p. Since by Lemma 3. because it implies d = 0. such that j = i ± ε (mod k ). The second case is impossible. a < k . Without loss of generality let j = i + ε (mod k ) (the proof follows similar lines if j = i − ε (mod k )). b)-code in S (n. thus extreme vertices would have in this case exactly a + 1 neighbors in the code. We extend C to C by setting i1 i2 . Then there exists (a. and one more additional neighbor ii − ε. First we observe that the vertex ii has 2p neighbors in C . The ﬁrst case implies n = 2. 1 ≤ ε ≤ p. and by the left side of the equality it is an integer. k ) that does not include any of the extreme vertices. Neighbors of ij in K (ij ) ∩ C are of the form il with l ≡ i ± ε (mod k ). Then ij has 2p neighbors in K (ij ) ∩ C and vertex ji C = {ij | j = i ± ε (mod k ). Then there exists ε.and (a.1. Moreover. Let ij ∈ / C . We distinguish two cases. b)-code that includes none of the extreme vertices. k ). We extend this code inductively to S (n.68 Laurent Beaudou. Proof: By Lemma 3. We claim that is an (a. We ﬁrst observe the following useful fact. k ) that does not include any of the extreme vertices. d = k and k = 2a + 1. k ) for all n ≥ 3. a + 2) In this section we construct (a. . k ) for all n ≥ 3 as follows. a)-code of S (n. k ) when at least one of a and k is even.

k ) as follows. First set ii ∈ C for all i = 0. a + 2)-code C in S (2. a)-code contains no extreme vertices we infer: Corollary 4. a)-code in S (2. . Then ij has precisely 2p + 1 = a neighbors in K (ij ) ∩ C and its neighbor ji does not belong to C . but since these will be clear from the context we write simply w(x).5 Let C be an (a. By Lemma 3. k ) is a near code . .3 Let C be an (a. . for j = i set ij ∈ C if and only if there is some t such that vt = i and vt+1 = j (where subscript are modulo a(2a + 1)). since i + q ≡ i − q ( mod k ).1. a + 1)-codes in Theorem 1. k ). 2a + 1 be its vertices. a + 1)-code To settle the case of (a.1. Given a set C of vertices of a graph G and a vertex x. Let ij ∈ C . k ). 5) using an Eulerian tour in K5 . Before we need some additional deﬁnitions. a + 2)-code. Note ﬁnally that in both cases C contains no extreme vertex. otherwise . k − 1. va(2a+1) be an Eulerian tour T in Q. Therefore C is an (a. Case 2. hence Lemma 4. . k ). Then Q is an Eulerian graph and let v1 . a)-code in S (2. x ∈ C. a + 2)-code in S (2. the edge ij of Q must belong in one way in T .1 applies. a = 2p + 1 and k = 2q . there must be some t such that vt = j and vt+1 = i. Proof: Let Q be the complete graph on k = 2a + 1 vertices and let 1. Then |C | = a · k n−1 . 2 where a (2. .Covering codes in Sierpi´ nski graphs 69 which is the neighbor of ij not belonging to the clique K (ij ) also does not belong to C . .4 S (2. 2a + 1) then all extreme vertices must belong to C moreover from C one may exhibit an Eulerian tour of K2a+1 (contract all vertices of C belonging to the same (2a + 1)-clique and consider the orientation of the edges between different cliques of S (2. be the weight of x. . where an edge is always oriented from a vertex in the code to the vertex not in the code). Then ij has 2p neighbors in K (ij ) ∩ C . k ). Combining Lemma 3.) Second. We claim that C = {ij | j = i ± ε (mod k ). Hence ij has precisely 2p + 1 = a neighbors in C . we prove a stronger result. a + 2)-codes in S (2. Therefore C is an (a. Then ij has a + 1 neighbors in C ∩ K (ij ) plus a neighbor ji. Since ij ∈ C . By deﬁnition of an Eulerian tour. a + 2)-code of S (2. Let ij ∈ / C . let ij ∈ C . w ( x) = a + 1 − |N (x) ∩ C |. . Note that ji ∈ C . Therefore C is an (a. 2a + 1). let a − | N ( x) ∩ C | . a < k and it follows 2p + 1 < 2q and therefore p < q . Then |C | = 2a2 + 3a + 1. a)-code in S (n. Moreover for vertex ji it follows that ji ∈ C . Using T we describe an (a. l = i + q (mod k )} is an (a.) Then we say that a subset C of vertices of S (n. So all (a. ✷ 5 Case of (a. 4)-code is constructed in S (2. 2. 2a + 1) are obtained by the construction given in the proof of Lemma 4. In addition.4. 2a + 1) contains an (a. 0 ≤ i ≤ k − 1. . k ). .3 with the fact that an (a. (The construction is illustrated in Fig. a)-code in S (n. 1 ≤ ε ≤ p} ∪ {il | 0 ≤ i ≤ k − 1. . . ✷ One may remark that if there is an (a. (The weight of x is also a function of C and a (and of G). Corollary 4. Lemma 4. 2a + 1). a + 2)-code C in S (2.

. in for all i2 . • W E n is a near code where n is even. We denote by • extreme vertices that lie in the near code and by ◦ the other extreme vertices. Now we are ready to state: .70 Laurent Beaudou. we have that C ∩ V (Si ) is a near code in Si if C is a near code in S (n. Furthermore. Let us now deﬁne the following special near codes C : • SOn is a near code where n is odd. Michel Mollard 04 00 01 41 03 02 14 44 40 43 42 30 31 24 10 11 12 4 0 1 34 20 13 3 2 33 32 23 22 21 Eulerian tour in K5 : 0→1→2→3→4→0→2→4→1→3→0 Fig. the other k − a extreme vertices are •∗ . the other k − a extreme vertices are ◦+ • SE n is a near code where n is even. a + 1 many extreme vertices are •∗ . a many extreme vertices are •+ . . k )) \ X (S (n. • W On is a near code where n is odd. . k ) induced by {ii2 . 5) obtained via an Eulerian tour in K5 if w(x) = 0 for all x ∈ V (S (n. a many extreme vertices are •+ . k − 1}} . . . . 4) code (black vertices) in S (2. the other k − a − 1 extreme vertices are ◦∗ . a + 1)-code. Given n and k . The graph Si is isomorphic to S (n − 1. we add the subscript ∗ for a vertex of weight 0 and the subscript + for weight 1. k ). . For example. ◦+ is an extreme vertex which is not in the code and which has a neighbors in the code. k ). the other k − a − 1 extreme vertices are ◦∗ Note that by deﬁnition SOn is an (a. . in ∈ {0. k ) (for short Si ) the subgraph of S (n. k )). 2: A (2. and w(x) ≤ 1 for vertices x in X (S (n. Matjaˇ z Kovˇ se. . a + 1 many extreme vertices are ◦+ . Sylvain Gravier. we denote Si (n. From the deﬁnition of the near code and since the only edges between Si and G − Si are incident to extreme vertices of Si . k )). Sandi Klavˇ zar.

Now for each extreme vertex y = x of S1 of kind •∗ we need an Si containing a ◦+ as extreme vertex. Similarly. k ) must be matched together in an arbitrary way. each remaining •+ not an extreme vertex of S (n. Suppose next that there is an extreme vertex x of kind •∗ . there are a additional graphs Si such that Si ∩ C is isomorphic to SE n−1 (altogether there a + 1 such graphs). there are a graphs Si such that Si ∩ C is isomorphic to W On−1 . Therefore. Now to obtain a near code.Covering codes in Sierpi´ nski graphs 71 Theorem 5. The k − a − 1 other Sj graphs contain a + 1 extreme vertices ◦+ . in particular C is an (a. k )). Now for each extreme vertex y = x of S1 of kind •+ we need an Si containing a •+ as extreme vertex. 0 ≤ i ≤ k − 1. for the •∗ extreme vertices of S1 . it is clear that any near code in the clique Kk is a set of a or a + 1 vertices which give respectively a W O1 and a SO1 code. To obtain a near code. k ) with n ≥ 2. k ) belonging to some Si must be •+ . If such a vertex is not extreme in S (n. . Further we divide induction step into two cases. a ≥ 0 and k > a be integers. Without loss of generality one may assume that x belongs to S1 . Similarly. Suppose that there is an extreme vertex x of the kind •∗ . If such a vertex is not extreme in S (n. are : •+ — •+ . The proof works by induction on n. If C contains a ◦+ the proof goes along similar lines as above. For n = 1. Therefore. it follows that each Sj must contain an extreme vertex ◦+ of S (n. we need k − a graphs Sj having extreme vertex of kind ◦+ which implies that Sj ∩ C is isomorphic to W E n−1 . k ) belonging to some Si must be •∗ . Now for each extreme vertex of S1 of kind •+ we need an Si containing a •+ as extreme vertex. for the ◦∗ extreme vertices of S1 . Therefore. according to the parity of n. Since there are a + 1 graphs Si containing a extreme vertices •+ and only one edge between two distinct Si . it implies that each extreme vertex of S (n. Without loss of generality one may assume that x belongs to S1 . The main argument of the proof is that the only possible matchings between extreme vertices of different subgraphs Si . Also the remaining vertices ◦∗ can be matched together where k − a − 1 of such vertices belong to X (S (n. hence they can be matched together. we need k − a − 1 graphs Sj having extreme vertex of kind ◦+ which by induction hypothesis implies that Sj ∩ C is isomorphic to W E n−1 . By induction hypothesis. Since there are a graphs Si containing a extreme vertices •+ and only one edge between two distinct Si . The near codes of S (n. Therefore C = SE n . Hence in this case we get C = W On . k ). Without loss of generality one may assume that x belongs to S1 . Since we have only (k − a) · a vertices of kind •∗ . for the •∗ extreme vertex y = x of S1 . By induction hypothesis. Suppose that there is an extreme vertex x of kind •+ . k ) it must be matched to a •∗ vertex. we need k − a − 1 graphs Sj having extreme vertex of kind ◦∗ which implies that Sj ∩ C is isomorphic to SOn−1 . C ∩ S1 is isomorphic to SE n−1 . k ). k )) ( and are from (a + 1) SE n−1 graphs) and the same number of ◦+ vertices . each remaining •+ not in an extreme vertex of S (n. Now. k ) must be matched together. C ∩ S1 is isomorphic to SE n−1 . k ) are precisely SOn and W On if n is odd and SE n and W E n if n is even. k ) it must be matched to a •∗ vertex. there are a − 1 graphs Si such that Si ∩ C is isomorphic to SE n−1 . By induction hypothesis. a + 1)-code in S (n. Proof: Let C be a near code of S (n. C ∩ S1 is isomorphic to SOn−1 . it implies that each extreme vertex of S (n. Assume ﬁrst that n is odd. ◦∗ — ◦∗ and •∗ — ◦+ . Similarly. Next let n be even and n ≥ 2.a + 1 of them in each of k − a − 1 copies of W E n−1 . If C contains a ◦∗ the arguments are similar.1 Let n ≥ 1. the k − a other Sj contain a + 1 extreme vertices ◦+ . We have exactly (k − a − 1) · (a + 1) vertices of kind •∗ that are not in X (S (n.

Without loss of generality one may assume that x belongs to S1 . 3) is illustrated in Fig. Corollary 5. b)-codes in graphs S (n.72 Laurent Beaudou.) Different types of extreme vertices of graphs S (n − 1. k ) are unique up to symmetries. k ) must be matched together. Similarly we can conclude that the remaining a + 1 vertices of X (S (n. 2)-code in S (3. the uniqueness (modulo obvious symmetries) has been proved in [14]. Hence we have completed the induction step and hence the proof. where the unique means that there are always the same k building blocks of graphs S (n − 1. b)-codes of graphs S (n. For the case of (0. k ). a + 1)-codes would be of interest. for the •∗ extreme vertex of S1 . k ). Sylvain Gravier. k ) might have. By induction hypothesis. b for which there exist (a. k ). Corollary 5. ✷ The recursive construction of an (1. It follows C = SE n . where any automorphism of S (n. b)-codes in graphs S (n. k )) are ◦∗ . (Graphs S (n − 1. As exhibited in our paper all existing (a. k ) belonging to some Si . are constructed in a unique way from (a . k ) must be matched together. . k ) are matched together in S (n. perfect codes. Similarly. A more precise description (using labels) of the constructed (a. It follows C = W E n . k ). a + 1)-code if and only if n is odd and 0 ≤ a ≤ k − 1. 6 Concluding remarks We have thus characterized integers a. k ) there are k − a − 1 vertices of type ◦+ . where n is an odd number.3 Let C be an (a. Now to obtain a near code. hence there are k − a such extreme vertices in S (n. k ) or their weak codes. k ). it follows that among extreme vertices of S (n. Suppose ﬁnally that there is an extreme vertex x of kind ◦∗ . It follows that also all existing (a. Altogether there are a extreme vertices •+ of S (n. Since we have k − a − 1 graphs of type SOn−1 and in each of those graphs the same number of vertices ◦+ . each of the remaining •+ vertices that are not extreme vertices of S (n. k ) in a unique way. different types of weak codes. C ∩ S1 is isomorphic to SOn−1 . as in the case of (a. k ) with corresponding (weak) codes. if C contains a ◦+ the proof works similarly. k ) permutes extreme vertices and the corresponding subgraphs S (n − 1. comparing to each other. Again. each remaining ◦+ not an extreme vertex of S (n. The remaining unmatched vertices from SOn−1 are all •∗ and there is precisely one from each component. Now for each extreme vertex y of S1 of kind ◦+ . that is. Sandi Klavˇ zar. When C contains a •+ the proof works similarly. there are k − a − 2 additional graphs Si such that Si ∩ C is isomorphic to SOn−1 (altogether there k − a − 1 such graphs). The next consequences are immediate. 1)-codes. Michel Mollard To obtain a near code. As shown in [15]. Since there are a graphs Si containing a extreme vertices •+ and only one edge between two distinct Si . k ) is isomorphic to the symmetric group Sym(k ). a + 1)-codes. b )-codes of graphs S (n − 1. k ). a + 1)-code in S (n.2 Graphs S (n. 3. It follows that each extreme vertex of S (n. we need a + 1 graphs Sj having extreme vertex of kind •∗ which by induction hypothesis implies that Sj ∩ C is isomorphic to W On−1 . Then |C | = (a + 1) · kn +1 k+1 . y = x. that is of type W On−1 must be •+ . k ) admits an (a. Matjaˇ z Kovˇ se. but altogether the set of all k graphs S (n − 1. k ) with the corresponding (weak) codes is as a set always the same. n ≥ 2. the automorphism group of a graph S (n. we need an Si containing a ◦+ as extreme vertex. Therefore.

Lytsin and A. Weighted covering and packings. Perfect codes in graphs. B 15 (1973) 289–296. Sandi Klavˇ zar and Matjaˇ z Kovˇ se were also supported in part by the Ministry of Science of Slovenia under the grant P1-0297. Discrete Math. Jr. Combin. Amsterdam. J. Nelson. Cull and I. Inform. Biggs. Elsevier. S. 268 (2003) 31–48. 1997. S. On multiple coverings of the inﬁnite rectangular grid with balls of constant radius. Lytsin and H.Covering codes in Sierpi´ nski graphs ∗ ∗ 73 SO1 ∗ ∗ ∗ SO1 ∗ ∗ W O1 + + + SE 2 ∗ + + + + + + + W O1 + W O1 + ∗ ∗ SO1 ∗ W O1 + + W E2 + ∗ + ∗ SE 2 ∗ ∗ ∗ SO1 ∗ SO1 ∗ ∗ ∗ Fig. [2] N. Acknowledgements This research was supported in part by the Slovene-French project Proteus 17934PB. Lobstein Covering Codes. Honkala. Error-correcting codes on the Towers of Hanoi graphs. Discrete Math.A. I. 3). 208/209 (1999) 157–175. References [1] M. [4] G. 3: Recursive construction of an (1. I. 2)-code in Sierpi´ nski graph S (3. Cohen.F. [3] G. . Cohen.. IEEE Trans. Mattson. Axenovich. [5] P. Theory 41 (1995) 1856–1967. Theory Ser. Honkala.

Perry. Dorbec. 57 (1998) 367–376. Pˇ r´ ırod. 5 (1998) 128–142. Michel Mollard [6] P. Bull. Des. Taiwanese J. Gravier.A. Codes Cryptogr. Slater. III 27(47) (1992) 343–364. 12 (2008) 513–522. Vˇ ed no. Discrete Math. Math. and P. P. Matjaˇ z Kovˇ se. Kratochv´ ıl. Mohar. [10] P. (In: T. Telle. 52 (2009) 209–218. T.T. Telle.L. 9 (2005) 671–681. [13] S. 10 (2003) 206–224. Lipscomb. Fiala. Hedetniemi.A. 47(122) (1997) 95–104. [11] M. Klavˇ zar. Taiwanese J. S. [12] S. Perfect codes on the Towers of Hanoi graphs.A. 2009. . Math. Completeness of the Lipscomb space. Hedetniemi. Graphs S (n. Nordic J. Comput. Comput. Nordic J. 1-perfect codes in Sierpi´ nski graphs. Crossing numbers of Sierpi´ nski-like graphs. Vˇ ed Rada Mat. Rozpravy Ceskoslovensk´ e Akad. Berlin. [19] S. Mollard.-K. Haynes. Soc. Heggernes. Sandi Klavˇ zar.C. Ser. k ) and a variant of the Tower of Hanoi problem.74 Laurent Beaudou. and total-colorings of Sierpi´ nski-like graphs. Math. Soc. Graph Theory 50 (2005) 186–198. Comput. Akademia Praha. Marcel Dekker. Heggernes and J. Nordic J. 1991. Soc. Complexity of domination-type problems in graphs. Vertex-. Milutinovi´ c. 7. [21] J. I. Coloring Sierpi´ nski graphs and Sierpi´ nski gasket graphs. Klavˇ zar. U. Proc. Jakovac and S. Springer-Verlag. Mollard. Domination in Graphs: Advanced Topics. 1)-labelings in Sierpi´ nski graphs. Mat. Nelson.) Chapter 9. 115 (1992) 1157–1165. Perfect Codes in General Graphs. 66 (2002) 369–384. Petr. Milutinovi´ c and C. [18] S. Austral. S. Austral. S. Milutinovi´ c. Klavˇ zar and B. Math. Lipscomb’s L(A) space fractalized in Hilbert’s l2 (A) space. Weighted codes in Lee metrics. A. W. Complexity results. 309 (2009) 1548–1556.A. P. New York. Telle Partitioning graphs into generalized dominating sets.L. Amer. 1998. [14] S. Klavˇ zar. edge-. Bull. 233–269. [20] U. Klavˇ zar and M. Glas. [7] J. J. Honkala and M. [9] S. Parks. Fractals and Universal Spaces in Dimension Theory. Klavˇ zar and U. [8] S. Generalized H -coloring and H -covering of trees. Kristiansen and J. [15] S. McRae and D. Sylvain Gravier. Codes and L(2. J. Li and I. [17] C. ˇ ˇ [16] J.A. Lipscomb and J. 1 (1994) 157–171. Gravier. J. Math. Czechoslovak Math.

Sign up to vote on this title

UsefulNot usefulCovering Codes in Sierpinski Graphs

Covering Codes in Sierpinski Graphs

- Graph Problem Sol
- Maximum index given domination number
- Artigo Original
- asst9 Math 239
- Hw6 Solutions
- Lecture 24
- maths
- Subgraph Iso Morph Ism in Planar Graphs And
- Breadth First Search
- applic
- Cs 180 Notes UCLA
- Assignment Solutionmcs031
- Week 5 Homework
- Basic Graph Algorithms
- 10305060
- Lect 8
- Roman Domination in Complementary Prism Graphs
- Cx 2424932496
- NP Complete
- Lecture_11_-_Graphs_P1_.pdf
- 2IL65 Assignment 7 Khanh
- 3.3 the Traveling-salesman Problem
- i(G)-Graph - G(i) Of Some Special Graphs
- Correctness Traversal
- Lecture 38
- Power Sol 98
- Algorithm Graphs with solution
- Simple Path Covers in Graphs
- Lecture-20-CS210-2012.pptx
- 04_Graphs1.pdf
- Covering Codes in Sierpinski Graphs